The ten-year Treasury yield is up about a full percentage point this year. Global stocks are up about 13 percent. Oil has been expensive enough, in a war year, that a growth model should have flinched. It barely did. Goldman Sachs now thinks the U.S. economy grew at a 3.4 percent pace in the third quarter, and it forecasts 2.2 percent growth in 2026 on a fourth-quarter-to-fourth-quarter basis. That is close to the bank’s 2.3 percent estimate of potential growth. It is only 0.3 percentage point below the forecast Goldman made in January, when it did not expect this rise in rates or this path for oil.
Abhay Duggiralas, on the bank’s economics team, asks the obvious question. Why has the economy been so resilient to these surprises? The answer in the note is not that rates do not matter. It is that they did not arrive alone. On Goldman’s financial-conditions index, the rise in interest rates since January has been nearly offset by higher equity prices, narrower credit spreads, and a weaker dollar. The net impulse to growth from that index is little changed from the start of the year. The bank puts the hit at about 0.1 percent of fourth-quarter-to-fourth-quarter growth in 2026.
This piece has one theme for investors. The resilience is an offset, not a law. Higher stock prices helped pay for higher yields. If the stock market stops paying, the offset stops, and the rate hit and the oil hit are left standing. The economy did not prove it can ignore a 100-basis-point rise in the ten-year. It proved that, so far, other markets moved the other way by about as much.
Nothing here is a forecast you should trade, and it is not advice to buy or sell stocks, bonds, or oil. A bank model is a set of offsets. Offsets reverse.
The gap that actually sets the price
Andrew Sheets, who runs fixed-income strategy at Morgan Stanley, framed the tension over the weekend in the bank’s Sunday Start note. Higher bond yields raise the return an investor can earn without owning a stock. In theory that raises the discount rate on future profits and lowers the value of equities. The theory has two moving parts. Sheets points at the Gordon growth model, which is simpler than it sounds. The value of a stock is the dividends you expect, divided by the gap between the return investors demand and the growth rate of those dividends. Call the demanded return r. Call growth g. The price is sensitive to r minus g, not to r by itself.
Raise r and the price falls, if g stays put. Raise g by as much as you raised r, or by more, and the price does not fall. Sheets thinks that is what a hot market does. Investors look at buoyant conditions and mark up the growth they expect, even faster than yields are rising. Psychology does the algebra. He thinks that is what has been happening. A year in which yields rise a percentage point and stocks rise 13 percent is not a violation of the model. It is a year in which the market decided g moved too.
That decision can be right. Earnings can arrive and justify the markup. It can also be a story people tell because the alternative is to cut the price. The test is not the model’s reputation. The test is whether the growth shows up outside the handful of companies that are doing the spending. On that test, the rest of the investment data is not hot at all.
The boom is a narrow boom
UBS, in a note last week, put a number on the narrowness. Outside technology, investment has effectively been in recession for at least two years. Private non-residential fixed investment has averaged just 0.1 percent a year over the last eight quarters. The long-run average is 3.2 percent. Residential investment has averaged minus 2.0 percent a year, against a long-run average of plus 2.2 percent. AI-related investment has grown 26 percent a year over those same eight quarters. The line UBS then draws is the uncomfortable one. With investment already so weak, the scope for rates to depress spending further may be limited.
Read that as a floor, not as health. A sector that is already flat or shrinking does not have much left to give a higher yield. The pain of the rate rise landed where borrowers are rate-sensitive, which is housing, and it landed earlier. What is left of “the economy” as a growth story is the consumer, plus a debt-funded build-out of computing power. Goldman’s own split of the year’s market moves says the same thing in a different dialect. Using the Federal Reserve’s FRB/US model, the bank finds only a modest effect on overall real GDP, a meaningful hit to residential investment, and slight boosts to consumption and to business investment. The chart of that exercise shows housing as the bar that drops. The other bars are small. The net bar is close to flat.
So the resilience slogan hides a three-layer cake. Housing has already taken the rate hit. Business investment outside the AI build has been near zero for two years. The AI build is large enough to hold the aggregate up, and it is funded by new debt in the hundreds of billions. Remove that layer, as the column wrapping these notes is eager to do, and the investment side of the U.S. economy does not look resilient. It looks tired. The slogan survives because people quote GDP, and GDP still adds the tired parts to the one hot part.
What the index says was cancelled
Goldman’s financial-conditions index is the cleanest picture of the cancel. Since January 1, the pieces have not moved together. The Treasury yield and the policy rate have tightened. Equity prices have eased conditions, because a higher stock market is a form of wealth and a cheaper cost of equity capital. Credit spreads have tightened, which is the opposite of stress. The dollar has weakened, which helps some activity and some earnings. The bank’s exhibit plots those contributions. The caption is the result. The overall index, and the growth impulse implied by it, have changed little. A note under the chart says the impulse assumes conditions stay flat after the date shown. January’s line and today’s line tell a similar story about 2026. The hit the bank is willing to put on fourth-quarter-to-fourth-quarter growth is 0.1 percent.
A tenth of a percent is not a boom and it is not a crisis. It is a remainder. Investors who say “look, five percent yields and the economy is fine” are looking at the remainder and calling it the whole sum. The whole sum included a stock market that went up, a credit market that did not crack, and a dollar that did not surge. Those are not laws of nature. They are this year’s other prices. Goldman is not claiming the rate rise was harmless in every sector. It is claiming the net, across the index, was small. Net is a word that should make you ask what is on the other side. On the other side is the equity price.
That is the link back to Sheets. If stocks are up because investors raised g, the higher stock price then feeds Goldman’s index and offsets the higher r. The story supports itself. Growth expectations lift prices. Prices loosen financial conditions. Looser conditions help the growth people expected. The loop is real until one turn fails. A failed turn looks like this. Earnings outside the AI complex do not arrive. Investors cut g. Prices fall. The equity contribution to the index flips from an offset into a tightening. The rate rise, which had been cancelled, is no longer cancelled. You do not need a new shock for that. You need the offset to go home.
Oil was a tax, and then a drilling story
Oil is the other surprise Goldman did not have in January. The bank’s rule of thumb is modest. A sustained $10 rise in the oil price lowers GDP growth by just over 0.1 percent, through consumption, because higher energy costs eat real purchasing power. Apply that rule to the path prices actually took, to an unusual widening in refined-product spreads, and to the bank’s own commodity forecasts through year-end, and the drag is larger. Goldman puts it at 0.4 percent of growth. That is the consumption hit before anyone counts the response of the oil industry.
The response was slow, and the note says why. Producers were reluctant to lift capital spending into a war with Iran and an oil outlook they did not trust. A higher price is not a plan if you think the price is a headline. The rig count changed the tone in the third quarter. Goldman’s chart, drawn from Baker Hughes, shows the number of active U.S. oil rigs falling through 2025 and then climbing back over the course of 2026. The bank reads the third-quarter rise as a pickup in energy capital spending. That spending is a plus for business investment. It claws back part of the consumption tax. Goldman’s net for the oil shock is a 0.3 percent hit to GDP growth. The fuel still costs households more than the new holes return to the total, but the gap is three-tenths of a point, not a recession by itself.
Put the two drags together. A tenth from financial conditions. Three-tenths from oil after the rigs wake up. Goldman’s summary is a combined drag of roughly 0.4 percent. The bank has cut its growth forecast this year, partly because of incoming data, by only a little less than that. The puzzle of resilience, on this arithmetic, is smaller than the dinner-table version. People expected a break. The models, once you include the offsets, expected a nick. The nick is about what the forecast gave up. What is left to explain is not a miracle. It is a small miss on the consumer.
The consumer is the leftover, and it is small
Goldman says the slight outperformance, relative to what the shocks implied, comes mostly from consumer spending. At the start of the year the bank forecast consumption growth of 2.4 percent on a fourth-quarter-to-fourth-quarter basis in 2026. Run the oil-shock model on the prices that actually happened and that forecast loses 0.6 percent. Higher equity prices give 0.2 percent back, through the wealth effect. The shocked forecast is 2.0 percent. The bank now thinks consumption will grow 2.2 percent, using incoming data that has been more solid than the model. The upside surprise is 0.2 percent.
Two-tenths of a point is a real number and a poor myth. It will not fund a claim that the household sector is immune to oil or to yields. It says households spent a little more than a model that already included a wealth boost from the stock market. The wealth boost is the fragile part. It is 0.2 percent of consumption growth that exists because equity prices went up. It is the same equity price that offset the rate rise inside the financial-conditions index. One market is doing two jobs. It is cancelling the bond selloff in the index. It is topping up consumption through the wealth effect. Those jobs end together if the market falls.
The column that carried the note draws the hard conclusion. If stocks finally sell off, the double hit of a market drop and a lost wealth effect means the United States will most likely slide into a recession. That sentence is the column’s, not a line I am putting in Goldman’s mouth. Goldman’s text, as quoted, stops at the arithmetic. A 0.2 percent consumption surprise. A 0.2 percent wealth offset. A 0.4 percent combined drag that roughly matches the forecast cut. The recession step is an inference. It is a fair inference to examine and a bad one to launder as the bank’s. A loss of the equity offset puts the rate hit and the oil hit back on an economy whose investment, outside AI, is already weak. That is a slower economy. Whether it is a recession depends on how far stocks fall, what credit does, and whether the AI spending itself is the thing that breaks. The note does not date that break. The column does not either. It only says the cushion is the market.
What an investor is actually holding
If you own the index, you are holding the offset. You are long the equity price that makes financial conditions look fine and that adds a wealth effect to consumption. You are also long the AI spending that is the only fast piece of investment. Those are related but not the same. The spending is a flow of cash into buildings, chips, and power. The price is what the market will pay for the earnings that flow is supposed to produce. Sheets’ g is the second one. UBS’s 26 percent is closer to the first. A flow can stay strong after a price falls. A price can stay high after a flow slows. The dangerous case is both at once. The flow slows because the debt funding it gets expensive, and the price falls because g is cut. Then the offset leaves and the growth story leaves in the same quarter.
If you own the ten-year, you are holding the rise that everyone says the economy survived. You were paid about 100 basis points more than at the start of the year, in Sheets’ tally. You were not paid that because the stock market agreed with you. The stock market went up anyway. Your yield is the r in someone else’s denominator. It hurts them only if their g does not rise with it. So far, in the aggregate, the market has acted as if g rose. Housing did not get to pretend. Residential investment is the sector Goldman’s model still shows as rate-sensitive, and UBS already has it shrinking. A bond investor who thinks the whole economy confirmed the yield is using GDP to hide a housing recession and a non-AI investment recession.
If you own oil producers, you are holding the slow half of Goldman’s oil math. The consumption tax hit first. The capex response waited until the rig count turned in the third quarter, and it waited because the war made the price hard to trust. A producer who did not drill in the spring was not asleep. The note says they did not want to spend into an outlook they could not underwrite. The autumn rig rise is a bet that the price will last. If it does not, the 0.1 percent of growth Goldman claws back from energy capex goes away, and the oil drag moves back toward the 0.4 percent consumption hit. Households would still have paid the expensive fuel. The offset from drilling would not stick.
What would end the offset
The theme fails if the offsets are actually fundamentals. Credit spreads can stay narrow because defaults stay low, not because buyers are complacent. The dollar can stay soft because the rest of the world is fine, not because it is doing U.S. growth a favor. Equity prices can stay high because earnings outside a slogan really do grow faster than the rise in r. In that world Sheets is right that g moved, and Goldman is right that the net financial-conditions hit is a tenth of a point, and the resilience is earned. You would see it in investment that is not only AI, and in a housing sector that stops shrinking once buyers adjust to the new yield. You are not seeing that yet. Eight quarters of 0.1 percent private investment growth, outside the technology boom, is a long time to call a transition.
The theme holds, and gets sharper, if stocks fall while yields stay up. Then r is high and g is being cut, which is the Gordon gap moving the wrong way. The financial-conditions index loses its equity offset and its spread offset if credit widens too. The wealth effect flips sign. Oil does not have to spike again for the hit to grow. The 0.3 percent oil drag is already in the model. What was hiding it was everything else. A investor who wants a checklist can keep it short. Is the equity price still doing the work. Are spreads still tight. Is the dollar still not tightening the index. Is AI spending still covering a private investment recession. If the first three turn and the fourth slows, the 2.2 percent growth forecast is a number from the offset year, not a number from the next one.
There is a quieter version that does not need a crash. UBS’s point is that rates cannot depress investment much further because investment outside technology is already depressed. That can be true and still be bad. A floor is not a launch. If AI spending merely stops accelerating, GDP growth can drift toward the potential rate Goldman already has, near 2.2 or 2.3 percent, without a recession and without a story of resilience. The word resilient would then be retired because there was nothing left to be resilient to. The rate rise would have done its work on housing and on ordinary capex, and the boom would have been a capex cycle with a ticker. That outcome is milder than the column’s recession line. It is also more consistent with the size of the numbers in the note. Tenths of a percent, not cliffs. The cliff is what you get if the market that supplied the tenths gives them back all at once.
The close
Yields rose about a percentage point. Stocks rose about 13 percent. Goldman’s 2026 growth forecast, fourth quarter to fourth quarter, is 2.2 percent, only a tenth or three below potential and only 0.3 point below a January call that did not include this year. The bank’s reason is an offset. Higher rates were cancelled, inside its financial-conditions index, by higher equities, tighter credit spreads, and a weaker dollar. The leftover hit is about 0.1 percent. Oil, after a slow rig response, is about 0.3 percent more. Together, roughly 0.4 percent, which is about how much the forecast has already been cut. The consumer is the leftover surprise, and it is 0.2 percent, half of it a wealth effect from the same stock market.
Under that aggregate, UBS sees a two-year investment recession outside technology, and a 26 percent pace inside the AI build. Housing is the sector the rate rise still hits. The economy people are calling resilient is a consumer, a wealth effect, and a narrow capex boom, standing on top of a private economy that already slowed down.
That is the theme. Do not confuse an offset with immunity. The stock market paid for the rate rise. It also paid a slice of consumption. If it stops paying, the yields and the oil tax are still there, and the growth that was supposed to outrun them has to show up without the help. Until it does, resilience is a description of this year’s other prices. It is not a promise about next year’s.
Important information
This article is for information and education only. It is not investment advice and not a recommendation to buy, sell, or hold any security, fund, or commodity. Growth forecasts miss. Financial conditions can change quickly. A model offset is not a floor under any price.
The Goldman figures are from the bank’s economics note as described on September 30, 2026, including comments attributed to Abhay Duggiralas. The Gordon growth framing is from Morgan Stanley’s Andrew Sheets in a Sunday Start note. The investment averages are from a UBS note cited the same week. The recession sentence attached to a stock selloff is the framing of the column that assembled these notes, not a quotation from Goldman’s forecast. Chart shapes are described only at the level the published captions support. This article does not consider any reader’s finances.